Distribution network topology identification is a key issue and an important prerequisite for distribution system operation. Because of the lack of sufficient measurement equipment and the frequent changes of distribution system topology, the actual distribution network topology identification problem is a highly challenging one. For this reason, this paper proposes a real-time topology identification method based on state estimation of synchronized measurement data. The proposed method can accurately estimate the topology of the distribution network using measurements from a small number of phasor measurement units. First, a structural equation model (SEM) is introduced in distribution networks to reveal the relationship between topology and branch currents. Then, a mixed integer programming (MIP) method is proposed for the topology identification problem. This decouples the imaginary and real parts of the complex equations, and linearizes the complex equations and constraints by means of Taylor expansion and equivalent relaxation. In addition, the distribution network operation mode constraints are reprogrammed on a node-by-node basis. Finally, the structural equation model-based mixed integer programming (SEM-MIP) method proposed can be solved using existing commercial solvers, and the effectiveness of the method has been verified by a real 12-bus distribution system and the IEEE 33-bus test system.